Harari’s Davos warning for schools: stop defining children by language performance

When AI outperforms humans in words, we need new signals of effort, growth, and ability.

Article link: https://open.substack.com/pub/johanosteyn/p/hararis-davos-warning-for-schools?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Audio summary: https://youtu.be/h5pTB5BCleA

Yuval Noah Harari’s recent Davos conversation on AI and humanity landed with me because it cuts through the classroom noise. We can debate which tools learners should use, which apps schools should ban, and which policies departments should publish. But Harari’s deeper point is that we are entering a world where “word outputs” are abundant, instant, and often convincing, and that changes what it means to educate.

If a machine can generate the essay, the summary, the poem, the motivation letter, and even the “reflection” in seconds, then the job of schooling cannot be to reward fluent text. It has to be to reward thinking you can evidence: reasoning, judgement, values, and the ability to explain how you got to an answer.

CONTEXT AND BACKGROUND

Harari’s Davos appearance was framed as a leadership issue, not a technical one, under the World Economic Forum’s 2026 programme on AI and humanity, featuring Harari in conversation with neuroscientist Irene Tracey.

His core claim is unsettling but practical: AI is not just a tool like a knife; it is an agent that can learn, change, decide, and manipulate. In his framing, anything made primarily of words becomes vulnerable because AI can assemble words at scale, and increasingly better than most humans. That lands directly in education, because so much of schooling is mediated through language: essays, tests, homework, memos, reports, and even “critical thinking” assignments that are, in practice, language performance.

South Africa adds a sharp edge to this. Inequality means the transition will not be evenly felt. Well-resourced schools can buy guidance, training, monitoring tools, and redesigned assessment practices. Under-resourced schools often experience disruption first and support later. Even the basic enabling layer matters: South African reporting has highlighted how education modernisation efforts depend on connectivity, devices, and reliable infrastructure, long before we get to sophisticated AI policy debates. And the political push to bring AI into schools is already visible in public commentary and government messaging.

INSIGHT AND ANALYSIS

Harari and Tracey circle a dilemma that many educators feel but struggle to name: if language is no longer scarce, we must stop treating language output as proof of understanding. That does not mean writing is worthless. It means writing needs to be reattached to accountable thinking. In other words, the classroom must shift from “submit the product” to “show the process”.

This is also where the children’s risk becomes real. Harari’s point is not only about cheating. It is about identity formation in a world where machines can simulate care, confidence, authority, and intimacy using words. Recent reporting and research coverage have raised concerns about teenagers’ interactions with chatbots, particularly when young people use them for sensitive conversations and guidance they might previously have sought from humans. You do not need to panic to see the governance gap: schools are being asked to educate in an environment where persuasive “voices” are increasingly non-human.

Assessment is the pressure point where this becomes practical. Internationally, institutions are experimenting with old ideas made newly relevant: oral exams, viva-style defences, in-class problem solving, and longer-term projects that generate artefacts over time. Even mainstream business coverage has started to notice the shift, as some educators explore oral assessment as a response to AI-written work. The direction is clear: the fewer marks you allocate to polish, and the more you allocate to reasoning you can interrogate, the less “instant text” breaks your system.

IMPLICATIONS

For business leaders and policymakers, the takeaway is straightforward: education reform cannot be reduced to procurement and platforms. The priority is assessment design and educator capability. If you want learners who can thrive in an AI-rich economy, you need to build systems that reward how people think, not how neatly they write.

For schools, a simple prioritisation approach helps. Start with high-risk assessment types (take-home essays, generic reflections, unsupervised online tests). Then introduce “process evidence” requirements: drafts, decision logs, source trails, and short oral check-ins. Increase the weight of tasks where learners must apply knowledge to a local context: case studies tied to their community, data they collect themselves, or projects that require iteration and critique.

For parents, the message is not “ban everything”. It is “stay present”. Ask children to explain, out loud, what they believe and why. Encourage learning habits that build a spine: reading, debating, making, testing, and reflecting in their own words. Harari’s warning is ultimately human-centred: if we define ourselves only by words, we will be outcompeted by systems that manufacture words. If we define ourselves by wisdom, judgment, and responsibility, we still have a future worth fighting for.

CLOSING TAKEAWAY

Harari’s Davos talk is not a prediction; it is a prompt for action. Education’s hardest pivot is to keep teaching thinking when words are cheap. In South Africa, the stakes are amplified by inequality: without a deliberate shift, the best-resourced learners will get guardrails and mentorship, while everyone else gets a flood of persuasive outputs and fragile assessment. The answer is not nostalgia for a pre-AI classroom. It is better design: assessments that demand reasoning, teachers supported to adapt, and a shared commitment to children’s development that cannot be outsourced to an agent, no matter how fluent it sounds.

Author Bio: Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net